A while ago, I was at the Grainger office in Jacksonville, and I noticed a wall showing a timeline of key historical events at Grainger. I noticed something called “Grainger’s Song” created back in the 1930s:
At this point, I’m sure you’re thinking what I was thinking. I wonder if I can generate a performance of this song using AI?
Idea 1: Find existing performance and swap in the Grainger’s Song lyrics
If you look at the top right of the image above, you’ll notice that Grainger’s Song was sung to the tune of “Might Lak a Rose”, so my first idea was to find an existing performance and swap in the Grainger’s Song lyrics.
Question: What does “Might Lak a Rose” actually sound like?
According to Wikipedia, “Mighty Lak’ a Rose” was created in 1901 with lyrics by Frank Lebby Stanton and music by Ethelbert Nevin. In other words, it’s in the public domain. No rights issues!
Wikipedia has a 1908 recording of a Lillian Nordica performance of “Mighty Lak’ a Rose” but it’s quite scratchy so I decided to search Youtube:
Not bad. I can imagine how the Grainger’s Song lyrics would have fit. As of right now, any performance before 1929 is public domain so I should be alright in terms of finding something.
Question: How does lyric replacement work?
Short answer, I’m not sure, but I did find some (pricey) paid services that would do it.
It made me think, is there an easier, simpler way to approach this?
Idea 2: Generate a song using Grainger’s Song lyrics
Enter Suno.
Prompt 1: “Make a song about the W.W. Grainger company”
That’s pretty good for an initial prompt. I was somewhat surprised at how relevant the lyrics were and could imagine this being used in a commercial.
But electric rock is definitely not the same genre as Mighty Lak’ a Rose.
Prompt 2: “A company song for W. W. Grainger using the tune of Mighty Lak’ a Rose”
Grainger's Anthem by @jchyip | Suno
sing-along folk song. Listen and make your own with Suno.
suno.com
Suno doesn’t seem to know Mighty Lak’ a Rose and I can’t imagine people singing along to it as a company song.
So far, I had only been using the basic prompt but Suno has a more advanced “custom” setting that allows you to specify the title, style of music, and lyrics.
Prompt 3: “A company song for W.W. Grainger” using the Grainger’s Song lyrics
Lyrics work. The genre doesn’t.
Looking at the metadata for some of the “Mighty Lak’ a Rose” Youtube videos, “dixieland” came up.
Prompt 4: “A company song for W.W. Grainger” using the Grainger’s Song lyrics and set style of music to “dixieland”
This is nice. And good enough that I didn’t think it was worth trying to do better… but I tried anyway with different prompts and styles of music, all of which ended up being worse.
Product Life Cycle and generative AI tools
When it comes to product development strategy, I often reference the product life cycle:
- Stage 1: Market Development is where you engage in experiments and innovation
- Stage 2: Growth is where you iterate on differentiation features
- Stage 3: Maturity is where you shift to “good enough” and start optimising for cost of operation
- Stage 4: Decline is where you look at how to efficiently exit.
Company songs don’t need to be differentiating.
Suno gets to “good enough” in minutes and gets even better with limited iteration. It makes me wonder about generative AI tools in general. A lot of activities and capabilities don’t need to be differentiating.
